REGEXP_SPLIT bug with diacritics

Trying to split the twitter stream sample, portuguese. Using:

SELECT text, regexp_split(text, ‘\Q \E’, ‘ALL’, 10) AS text_1
FROM (
SELECT text
FROM “twitter stream”.“twitter-stream.json”
WHERE lang = ‘pt’
) nested_0

But result is wrong when diacritic is present:

“bom diaaaaaa” [ “bom”, “diaaaaaa” ] OK

“Esse frio tá ótimo
Não disse pra que” [ “Esse”, “frio”, “t�”, " óti", “o \nN”, “�o di”, "se ", "ra " ] NOK